Polysemy Chinese entity relation identification method based on known network

The invention discloses a polysemy Chinese entity relationship recognition method based on a known network, and the method comprises the following steps: 1) performing vectorization of word granularity on each corpus sample in Chinese network event data based on the known network to obtain a word granularity vector of each word pair, then coding the position information of the granularity vector of each word, and obtaining a relative position code of each word in the corpus and a pre-labeled entity relation pair to be recognized; 2) generating a word granularity semantic vector set of each corpus sample according to a result obtained in the step 1); 3) generating a word granularity semantic vector set of each corpus based on the known network; 4) using each semantic vector and the corresponding position code to train a deep self-attention neural network to obtain a deep self-attention neural network encoder; 5) generating semantic vectors of words and vocabularies in the to-be-processed corpus and correspondi